Methodological workshop

Frequentist and Bayesian approaches to improving your statistical inferences
21 aprile 2015
April 21, 2015
Contatti: 
Segreteria Dipartimento di Psicologia e Scienze Cognitive
corso Bettini, 84 - Rovereto (TN)
Tel. 
+39 0464 808450 - 808413 - 8608 - 8610

Venue

Conference Room – Department of Psychology and Cognitive Science - Palazzo Fedrigotti - corso Bettini, 31 - Rovereto (Tn)

Time

 9.00 - 17.00

Speakers

  • Daniel Lakens- Eindhoven University
  • Luigi Lombardi - University of Trento

Scientific Coordinators

  • Luigi Lombardi & Maria Paola Paladino

Workshop description

In the first part of this workshop, a practical introduction is provided to recently developed statistical tools that can be used to deal with the inherent uncertainties in an inductive science. The goal is to allow researchers to improve the ways in which they design and evaluate research. The benefits of meta-analytic techniques such as p-curve analysis and meta-regression will be highlighted, with a focus on using these techniques to control for the effects of publication bias when evaluating studies. In addition, sequential analyses will be discussed, which will allow researchers to design well-powered studies by collecting data, analyzing it, collecting more data, and analyzing it, without p-hacking, even when effect sizes are uncertain. These techniques allow researchers to improve their statistical inferences from a Frequentist approach to statistics.

In the second part of the workshop, a gentle introduction is presented to some Bayesian perspectives in applied data analysis which constitute natural alternatives to standard null-hypothesis significance testing. Here the main objective is to provide some simple and manageable examples which highlight basic differences with respect to the frequentist perspective and stress the importance of shifting towards a more modeling oriented approach in data analysis. However, we will also see that there are some relevant differences in the way researchers adopt the Bayesian methodologies in applied data analysis and that they do not necessarily converge to the same substantive interpretations and goals.

For more details see here